This tool helps identify which prediabetes cluster (or subtype) you belong to, based on your clinical and metabolic profile. This tool is grounded in findings from our published research. It follows a precision medicine approach to assess better the risk of progression to type 2 diabetes and related conditions, such as liver and kidney diseases.

Note: This tool is intended for scientific and educational purposes only. It does not provide a clinical diagnosis or medical advice.

Objective

The primary goal is to classify individuals with prediabetes into distinct subphenotypes to support early risk assessment and personalized intervention strategies.

Scientific Background

This tool was developed using machine learning (XGBoost algorithm) trained on a cohort of individuals with prediabetes. The clustering approach was inspired by recent advancements in precision diabetes research, allowing identification of hidden subgroups based on clinical and biochemical profiles.

How the Clusters Were Identified

Clusters were derived using De Novo unsupervised learning methods applied to variables listed below. Each cluster reflects a distinct risk profile for prediabetes and complications.

• OGTT glucose values (fasting, 1-hour and 2-hour)
• HbA1c
• Triglycerides
• HDL cholesterol
• LDL cholesterol
• Waist circumference
• Age

Identified Prediabetes Clusters

• Cluster 1 – Young Overweight: Younger individuals with mild metabolic disturbances.
• Cluster 2 – Isolated IFG: Characterized by impaired fasting glucose with the lowest diabetes risk.
• Cluster 3 – Age-Related: Predominantly older individuals with moderate risk factors.
• Cluster 4 – High Glycemia: Highest glucose and lipid levels; highest risk of diabetes progression.
• Cluster 5 – Obesity and Dyslipidemia: Marked by obesity and abnormal lipid profiles, with elevated risk.

Potential Applications

• Research studies in metabolic phenotyping
• Targeted screening and prevention programs
• Educational tool for clinicians and students
• Basis for personalized intervention trials

Data Privacy

This tool does not store any personal data. All computations are performed locally in your browser session.

1. Upload CSV file only.
2. Ensure the dataset contains columns names in this order:
AGE, WAIST, HbA1C, TG, HDL, LDL, FBS, ogtt_1h, ogtt_2h
3. If your glucose values (FBS, OGTT) are in mg/dL, tick the checkbox to convert them to mmol/L. Model expects glucose in mmol/L.
4. Click 'Predict' to generate cluster labels.
5. View the subtype distribution and preview the results below.
6. Download your results.